Re: Enterprise Manager

From: Chuck <chuckh_nospam_at_softhome.net>
Date: 11 May 2004 17:18:23 GMT
Message-ID: <Xns94E6875C54024chuckhsofthomenet@130.133.1.4>

Holger Baer <holger.baer_at_science-computing.de> wrote in news:c7q15u$dm2$1_at_news.BelWue.DE:

> Chuck wrote:

>> I'm currently using OEM 9.2 (console and agents) and thinking about 
>> upgrading to a newer version. One thing I've always disliked is it's
>> poor ability to  monitor databases in an HACMP configuration. We have
>> several that fail over from one node to another. The only way we've
>> ever gotten the monitoring to work is to register each of these
>> database twice, once on each node. Does OEM 10 handle this any
>> better? If so, how? 

>
> I've got to admit my ignorance on the fact that I don't know what a
> HACMP configuration is, but if it in any way resembles the Microsoft
> Cluster Server (your description sounds like it) then all you have to
> do is to obtain a virtual ip-adress that the intelligent agent can
> use.
>
> Now you just have to make sure that the Intelligent Agent is failing
> over together with the database and the IP-adress and now you can
> monitor the database no matter on which node it is.
>
> The setup with Oracle Failsafe is quite easy (although only available
> on Windows) but not hassle free.
>
> I don't know if it changed with 10g, but then, I don't see the need
> for a change given an appropriate configuration.
>
> HTH
>
> Holger
>

Yes it's similar to what you're describing but it's not quite that simple. In OEM 9.2, the database is tied to a specific node. You might be able to get by with what you've described if there is only one database and one virtual address. But I have 12 databases and 12 virtual addresses all of which fail over independently. Any database could be running on any one node of the cluster at any time.

OIA will not discover virtual addresses. It discovers the fixed address/hostname of the node. Once a database fails over from the node it as discovered on, OEM can no longer be used to monitor or administer it. The problem is in OEM's assumption that a database is associated with one node. It is not. In this case it's associated with a *virtual* node.

The only solution to this problem so far has been to define each database multiple times, once on each node (and this is no small task). OEM always thinks one is up and the rest are down. It doesn't understand that they are the same db. OEM needs to divorce that DB from a specific physical node which as of 9i hasn't happened. I was hoping that either (a) this has been done in 10g, or (b) someone has a better way of handling this situation.

-- 
Chuck
Remove "_nospam" to reply by email

Received on Tue May 11 2004 - 12:18:23 CDT